CN112953973B - Hybrid attack detection method for continuous variable quantum key distribution system - Google Patents

Abstract

The invention discloses a hybrid attack detection method for a continuous variable quantum key distribution system, which comprises the steps of constructing the continuous variable quantum key distribution attack detection system; acquiring continuous variable quantum key communication data under different attack modes by adopting a built system, and processing and dividing a sample set; constructing a continuous variable quantum key distribution attack detection model based on multi-label learning and training to obtain an attack detection model; and monitoring actual communication by adopting an attack detection model. The hybrid attack detection method for the continuous variable quantum key distribution system provided by the invention learns and identifies the attack mode by adopting the sequencing support vector machine algorithm in multi-label learning, so that the method can accurately detect the attack types included in the hybrid attack for the quantum key distribution system, and has the advantages of high reliability, good practicability and wide application range.

Description

Hybrid attack detection method for continuous variable quantum key distribution system

technical field

The invention belongs to the field of quantum communication, in particular to a hybrid attack detection method for a continuous variable quantum key distribution system.

Background technique

Based on the information theory of the basic laws of quantum physics, quantum key distribution has the advantage of absolute security in theory, and is one of the most important applications of quantum technology. According to different carriers, quantum key distribution can be divided into discrete variable quantum key distribution (DVQKD) and continuous variable quantum key distribution (CVQKD). Compared with DVQKD, CVQKD started later, but has a higher key rate, so the application of continuous variable quantum key (CVQKD) is more widespread and popular. The Gaussian Modulated Coherent State (GMCS) protocol is currently the easiest to understand in terms of security and implementation, and has been theoretically proven to be resistant to collective attacks. However, some practical attack strategies can also undermine the security of GMCS CVQKD in real situations, such as Trojan horse attack, wavelength attack, calibration attack, local oscillator strength attack, saturation attack and homodyne detection blinding attack.

In view of the above situation, most of the current strategies are to add a suitable real-time monitoring module to the system; however, the real-time monitoring module can only prevent a single attack. At the same time, due to the defects of the actual device, the legitimate two parties must perform multiple iterative calculations to obtain an accurate estimated value, and after the key transmission is completed, the attack cannot be accurately detected when the attacker (Eve) attacks. In addition, some scholars have proposed a general attack detection scheme, which can resist as many attack types as possible, but this scheme can only detect one kind of attack. In real operation, the attacker Eve will not implement only one kind of attack, or there may be multiple attacker Eves to carry out different attacks at the same time, which makes this method unpractical.

SUMMARY OF THE INVENTION

The purpose of the present invention is to provide a hybrid attack detection method for a continuous variable quantum key distribution system with high reliability, good practicability and wide application range.

The hybrid attack detection method for the continuous variable quantum key distribution system provided by the present invention includes the following steps:

S1. Build a continuous variable quantum key distribution attack detection system;

S2. Use the continuous variable quantum key distribution attack detection system built in step S1 to obtain continuous variable quantum key communication data under different attack modes;

S3. Data processing is performed on the communication data obtained in step S2, and a sample set is divided;

S4. Construct a continuous variable quantum key distribution attack detection model based on multi-label learning, and use the sample set obtained in step S3 for training to obtain an attack detection model;

S5. Using the attack detection model obtained in step S4, the communication process of the actual continuous variable quantum key distribution system is monitored, so as to realize hybrid attack detection for the continuous variable quantum key distribution system.

The continuous variable quantum key distribution attack detection system described in step S1 includes a pulse laser at the sending end, a beam splitter at the sending end, a phase modulator at the sending end, an amplitude modulator at the sending end, a polarization beam splitter at the sending end, and a polarization beam splitter at the receiving end receiver, receiver first beam splitter, receiver second beam splitter, receiver phase modulator, receiver amplitude modulator, receiver light rate meter, receiver synchronization clock, receiver homodyne detector, receiver control The sending end pulse laser, the sending end beam splitter, the sending end phase modulator, the sending end amplitude modulator and the sending end polarization beam splitter are connected in series in sequence; the sending end pulse laser is used to generate optical pulses and send them to the sending end splitter Beam splitter; the beam splitter at the sending end is used to split the received optical pulse into signal light and local oscillator light according to 10:90, and send the signal light to the phase modulator at the sending end, and send the local oscillator light to the sending end for polarization Beam splitter; the phase modulator at the transmitting end is used to phase modulate the received signal light before sending it to the amplitude modulator at the transmitting end; the amplitude modulator at the transmitting end is used to perform amplitude modulation on the received optical signal before sending it to the transmitting end for polarization Beam splitter; the polarizing beam splitter at the transmitting end is used to transmit the received local oscillator light and the signal light modulated by phase and amplitude to the receiving end after time division and polarization multiplexing; The signal is decomposed into signal light and local oscillator light, and the signal light is sent to the amplitude modulator at the receiving end, and the local oscillator light is sent to the first beam splitter at the receiving end; the first beam splitter at the receiving end is used to convert the received local oscillator light The signal is split by 10:90 and sent to the second beam splitter at the receiving end and the phase modulator at the receiving end; Two beams are sent to the optical rate meter at the receiving end and the synchronization clock at the receiving end respectively; the phase modulator at the receiving end is used to phase modulate the received 10% beam of the local oscillator optical signal and send it to the homodyne detector at the receiving end; The amplitude modulator at the receiving end is used to attenuate the received signal light with the maximum probability of the set value, and then perform real-time shot noise estimation, and send the result to the homodyne detector at the receiving end; the light rate meter at the receiving end is used according to the The received signal is subjected to optical power detection, and the result is uploaded to the receiver controller; the receiver synchronous clock is used to generate a clock signal according to the received signal, and the result is uploaded to the receiver controller; the receiver homodyne detector is used for It is used to perform homodyne detection on the received local oscillator light and signal light, and upload the results to the receiver controller; the receiver controller is used for sampling and attack detection according to the received signal.

The obtaining of the continuous variable quantum key communication data under different attack modes described in step S2 is specifically to obtain the continuous variable quantum key communication data during normal communication and the continuous variable quantum key communication data when attacked; the described Attacks include single-mode or combined-mode attacks in calibration attacks, low-intensity attacks, intercept-replay attacks, and saturation attacks.

I_LOi为第i组测量数据中的本振光的强度，N_0i为第i组测量数据中的散粒噪声方差，

为第i组测量数据中的接收端测量到的正交平均值，V_ui为第i组测量数据中的接收端测量到的正交方差；q种可能的攻击类型标记为多标签空间Y＝{y₁,y₂,...,y_q}，y的取值为0或1，0代表未遭受该攻击，1表示遭受该攻击；构建样本数据D＝{(x_i,y_i)|1≤i≤n}；然后将样本数据采用最大最小归一化算法进行数据归一化处理，最后按照设定的比例划分为训练集和测试集。In step S3, data processing is performed on the communication data obtained in step S2, and the sample set is divided, specifically, the continuous variable quantum key communication data in the normal communication mode and in the attacked mode, and the measurement data are marked as d-dimensional features. X={x ₁ ,x ₂ ,...,x _d }, where

I _LOi is the intensity of the local oscillator light in the i-th group of measurement data, N _0i is the shot noise variance in the i-th group of measurement data,

is the quadrature mean value measured by the receiver in the i-th group of measurement data, V _ui is the quadrature variance measured by the receiver in the i-th group of measurement data; q possible attack types are marked as multi-label space Y= {y ₁ , y ₂ ,...,y _q }, the value of y is 0 or 1, 0 means not being attacked, 1 means being attacked; constructing sample data D={(x _i ,y _i ) |1≤i≤n}; then the sample data is normalized by the maximum and minimum normalization algorithm, and finally divided into training set and test set according to the set ratio.

The construction of the continuous variable quantum key distribution attack detection model based on multi-label learning described in step S4 is specifically to use the following steps to build the model:

A. The internal structure of the model is a sorting support vector machine of multi-label algorithm, including training linear model and training threshold function;

B. The training linear model is specifically:

In the label, the judgment is based on whether it is relevant or irrelevant; the following formula is used to express the classification hyperplane corresponding to the relevant label y _j and the irrelevant label y _k :

<ω _j -ω _k ,x>+b _j -b _k =0

where ω _j is the weight of label y _j ; ω _k is the weight of label y _k ; b _k is the offset of label y _k ; b _j is the offset of label y _j ;

The optimization problem corresponding to the algorithm in the real situation is represented by the following formula: the summation of maximizing the decision edge and minimizing the Ranking loss function:

Constraints: <ω _j -ω _k , ^xi ＞+b _j -b _k ≥1-ξ _ijk

为松弛变量集合；C为惩罚因子；where _ξijk is the slack variable,

is the set of slack variables; C is the penalty factor;

C. The training threshold function is specifically:

The threshold is expressed by the following formula:

为标签空间的补集；t为设定阈值；Where f _k ( _xi )=＜ω _k , _xi >+b _k , f(·) returns a real value, which represents the output value of the multi-label classification system on each label; Y is the label space;

is the complement of the label space; t is the set threshold;

D. Based on the training linear model and the training threshold function, the final multi-label classifier is obtained as h(x)={y _k |f _k (x)>t(x), 1≤k≤q}; q is possible Number of attack types; t( ) is the threshold function used in step C; h(x) is the final output of 0, 1 strings, corresponding to whether each attack is received.

The hybrid attack detection method for the continuous variable quantum key distribution system provided by the present invention uses the sorting support vector machine algorithm in multi-label learning to learn and identify the attack mode, thereby ensuring that the method of the present invention can accurately detect The attack type included in the hybrid attack against the quantum key distribution system is obtained, and the method of the invention has high reliability, good practicability and wide application range.

Description of drawings

FIG. 1 is a schematic flow chart of the method of the present invention.

FIG. 2 is a functional block diagram of the continuous variable quantum key distribution attack detection system in the method of the present invention.

Detailed ways

1 is a schematic flow chart of the method of the present invention: this hybrid attack detection method for a continuous variable quantum key distribution system provided by the present invention includes the following steps:

S1. Build a continuous variable quantum key distribution attack detection system;

S2. Use the continuous variable quantum key distribution attack detection system built in step S1 to obtain continuous variable quantum key communication data under different attack modes; specifically, obtain the continuous variable quantum key communication data during normal communication, and when attacked continuous variable quantum key communication data; the attacks include calibration attack, low-intensity attack, interception-retransmission attack and saturation attack in a single mode attack or combined mode attack, a total of 15 attack modes;

I_LOi为第i组测量数据中的本振光的强度，N_0i为第i组测量数据中的散粒噪声方差，

为第i组测量数据中的接收端测量到的正交平均值，V_ui为第i组测量数据中的接收端测量到的正交方差；q种可能的攻击类型标记为多标签空间Y＝{y₁,y₂,...,y_q}，y的取值为0或1，0代表未遭受该攻击，1表示遭受该攻击；构建样本数据D＝{(x_i,y_i)|1≤i≤n}；然后将样本数据采用最大最小归一化算法进行数据归一化处理，最后按照设定的比例划分为训练集和测试集；S3. Perform data processing on the communication data obtained in step S2, and divide the sample set; specifically, the continuous variable quantum key communication data in the normal communication mode and in the attacked mode, the measurement data is marked as the d-dimensional feature X={ x ₁ ,x ₂ ,...,x _d }, where

I _LOi is the intensity of the local oscillator light in the i-th group of measurement data, N _0i is the shot noise variance in the i-th group of measurement data,

is the quadrature mean value measured by the receiver in the i-th group of measurement data, V _ui is the quadrature variance measured by the receiver in the i-th group of measurement data; q possible attack types are marked as multi-label space Y= {y ₁ , y ₂ ,...,y _q }, the value of y is 0 or 1, 0 means not being attacked, 1 means being attacked; constructing sample data D={(x _i ,y _i ) |1≤i≤n}; then the sample data is normalized by the maximum and minimum normalization algorithm, and finally divided into training set and test set according to the set ratio;

S4. Build a continuous variable quantum key distribution attack detection model based on multi-label learning, and use the sample set obtained in step S3 for training to obtain an attack detection model; specifically, the following steps are used to build the model:

A. The internal structure of the model is a sorting support vector machine of multi-label algorithm, including training linear model and training threshold function;

B. The training linear model is specifically:

In the label, the judgment is based on whether it is relevant or irrelevant; the following formula is used to express the classification hyperplane corresponding to the relevant label y _k and the irrelevant label y _l :

<ω _j -ω _k ,x>+b _j -b _k =0

where ω _j is the weight of label y _j ; ω _k is the weight of label y _k ; b _k is the offset of label y _k ; b _j is the offset of label y _j ;

The optimization problem corresponding to the algorithm in the real situation is represented by the following formula: the summation of maximizing the decision edge and minimizing the Ranking loss function:

Constraints: <ω _j -ω _k , ^xi ＞+b _j -b _k ≥1-ξ _ijk

为松弛变量集合；C为惩罚因子；where _ξijk is the slack variable,

is the set of slack variables; C is the penalty factor;

C. The training threshold function is specifically:

The threshold is expressed by the following formula:

为标签空间的补集；t为设定阈值；Where f _k ( _xi )=＜ω _k , _xi >+b _k , f(·) returns a real value, which represents the output value of the multi-label classification system on each label; Y is the label space;

is the complement of the label space; t is the set threshold;

D. Based on the training linear model and the training threshold function, the final multi-label classifier is obtained as h(x)={y _k |f _k (x)>t(x), 1≤k≤q}; q is possible Number of attack types; t( ) is the threshold function used in step C; h(x) is the final output of 0, 1 strings, corresponding to whether each attack is received.

S5. Using the attack detection model obtained in step S4, the communication process of the actual continuous variable quantum key distribution system is monitored, so as to realize hybrid attack detection for the continuous variable quantum key distribution system.

As shown in FIG. 2, it is the continuous variable quantum key distribution attack detection system described in step S1 in the method of the present invention; the system specifically includes a sending end pulse laser, a sending end beam splitter, a sending end phase modulator, a sending end Amplitude modulator at the transmit end, polarization beam splitter at the transmit end, polarization beam splitter at the receive end, first beam splitter at the receive end, second beam splitter at the receive end, phase modulator at the receive end, amplitude modulator at the receive end, light rate meter at the receive end , the receiving end synchronous clock, the receiving end homodyne detector, the receiving end controller; the sending end pulse laser, the sending end beam splitter, the sending end phase modulator, the sending end amplitude modulator and the sending end polarization beam splitter are serially connected in sequence; The sending-end pulse laser is used to generate optical pulses and send them to the sending-end beam splitter; the sending-end beam splitter is used to split the received optical pulses into signal light and local oscillator light according to 10:90, and send the signal light. to the phase modulator at the sending end, and send the local oscillator light to the polarization beam splitter at the sending end; the phase modulator at the sending end is used to phase modulate the received signal light, and then send it to the amplitude modulator at the sending end; the amplitude modulator at the sending end is used for The received optical signal is amplitude modulated and then sent to the polarizing beam splitter at the transmitting end; the polarizing beam splitter at the transmitting end is used to time-division and polarization-multiplex the received local oscillator light and the signal light that has undergone phase and amplitude modulation. , sent to the receiving end; the receiving end polarization beam splitter decomposes the received signal into signal light and local oscillator light, and sends the signal light to the receiving end amplitude modulator, and sends the local oscillator light to the receiving end first beam splitter ; The first beam splitter at the receiving end is used to split the received local oscillator optical signal according to 10:90 and transmit it to the second beam splitter at the receiving end and the phase modulator at the receiving end respectively; the second beam splitter at the receiving end is used for Divide the received 90% of the local oscillator optical signal into two beams equally, and send them to the receiving end light rate meter and the receiving end synchronous clock respectively; the receiving end phase modulator is used to The vibrating light signal is phase-modulated and sent to the homodyne detector at the receiving end; the amplitude modulator at the receiving end is used to attenuate the received signal light with the maximum probability of the set value, and then perform real-time shot noise estimation and send the result. To the receiving end homodyne detector; the receiving end light rate meter is used to detect the optical power according to the received signal, and upload the result to the receiving end controller; the receiving end synchronization clock is used to generate the clock signal according to the received signal, And upload the results to the receiver controller; the receiver homodyne detector is used to perform homodyne detection on the received local oscillator light and signal light, and upload the results to the receiver controller; the receiver controller is used to The signal is sampled and attack detected.

Claims (3)

1. A hybrid attack detection method aiming at a continuous variable quantum key distribution system comprises the following steps:

s1, building a continuous variable quantum key distribution attack detection system;

s2, acquiring continuous variable quantum key communication data in different attack modes by adopting the continuous variable quantum key distribution attack detection system established in the step S1;

s3, carrying out data processing on the communication data obtained in the step S2, and dividing a sample set; specifically, the method marks continuous variable quantum key communication data in a normal communication mode and an attack mode as d-dimensional characteristic X ═ X₁,x₂,...,x_dTherein of

I_LOiFor the intensity of the local oscillator light, N, in the ith set of measurement data_0iFor shot noise variance in the ith set of measurement data,

is the orthogonal average value, V, measured by the receiving end in the ith group of measured data_uiThe orthogonal variance measured by a receiving end in the ith group of measurement data; the q possible attack types are marked as multi-label space Y ═ Y₁,y₂,...,y_qY is 0 or 1, 0 represents that the attack is not suffered, 1 tableIndicating that the attack is suffered; constructing sample data D { (x)_i,y_i) I is more than or equal to 1 and less than or equal to n; then, carrying out data normalization processing on the sample data by adopting a maximum and minimum normalization algorithm, and finally dividing the sample data into a training set and a test set according to a set proportion;

s4, constructing a continuous variable quantum key distribution attack detection model based on multi-label learning, and training by adopting the sample set obtained in the step S3 to obtain an attack detection model; specifically, the model is constructed by adopting the following steps:

A. the internal structure of the model is a sequencing support vector machine of a multi-label algorithm, and the sequencing support vector machine comprises a training linear model and a training threshold function;

B. the training linear model is specifically as follows:

in the label, the related or unrelated is used as a judgment; the related label y is expressed by the following formula_kAnd irrelevant label y_lCorresponding classification hyperplane:

＜ω_j-ω_k,x＞+b_j-b_k＝0

in the formula of omega_jAs a label y_jThe weight of (c); omega_kAs a label y_kThe weight of (c); b_kAs a label y_kThe offset of (2); b_jAs a label y_jThe offset of (2);

the optimization problem corresponding to the algorithm under the real condition is expressed by the following formula: the sum of the maximized decision edge and the minimized Ranking loss function:

constraint conditions are as follows: < omega_j-ω_k,x_i＞+b_j-b_k≥1-ξ_ijk

ξ_ijk＞0,1≤i≤n,

Xi in the formula_ijkIn order to be a function of the relaxation variable,

is a relaxation variable set; c is a penalty factor;

C. the training threshold function is specifically:

the threshold value is expressed by the following equation:

wherein f is_k(x_i)＝＜ω_k,x_i＞+b_kF (-) returns a real value representing the output value of the multi-label classification system on each label; y is a label space;

is the complement of the label space; t is a set threshold;

D. based on a training linear model and a training threshold function, obtaining a final multi-label classifier h (x) ═ y_k|f_k(x) Q is more than t (x) and is more than or equal to 1 and less than or equal to k; q is the number of possible attack types; t (-) is the threshold function employed; h (x) finally outputting character strings of 0 and 1, and correspondingly judging whether the attack is received or not;

and S5, monitoring the communication process of the actual continuous variable quantum key distribution system by adopting the attack detection model obtained in the step S4, thereby realizing the hybrid attack detection aiming at the continuous variable quantum key distribution system.

2. The hybrid attack detection method for the CVQKD (continuous variable quantum key distribution) system as claimed in claim 1, wherein the CVQKD attack detection system of step S1 comprises a transmitting-end pulse laser, a transmitting-end beam splitter, a transmitting-end phase modulator, a transmitting-end amplitude modulator, a transmitting-end polarization beam splitter, a receiving-end first beam splitter, a receiving-end second beam splitter, a receiving-end phase modulator, a receiving-end amplitude modulator, a receiving-end light rate meter, a receiving-end synchronous clock, a receiving-end homodyne detector and a receiving-end controller; the transmitting end pulse laser, the transmitting end beam splitter, the transmitting end phase modulator, the transmitting end amplitude modulator and the transmitting end polarization beam splitter are sequentially connected in series; the transmitting end pulse laser is used for generating optical pulses and transmitting the optical pulses to the transmitting end beam splitter; the transmitting end beam splitter is used for splitting the received optical pulse into signal light and local oscillator light according to the ratio of 10:90, transmitting the signal light to the transmitting end phase modulator and transmitting the local oscillator light to the transmitting end polarization beam splitter; the sending end phase modulator is used for carrying out phase modulation on the received signal light and then sending the signal light to the sending end amplitude modulator; the transmitting end amplitude modulator is used for carrying out amplitude modulation on the received optical signal and then transmitting the optical signal to the transmitting end polarization beam splitter; the polarization beam splitter at the sending end is used for transmitting the received local oscillation light and the signal light subjected to phase and amplitude modulation to a receiving end after time division and polarization multiplexing; the receiving end polarization beam splitter splits a received signal into signal light and local oscillator light, sends the signal light to a receiving end amplitude modulator, and sends the local oscillator light to a receiving end first beam splitter; the receiving end first beam splitter is used for splitting the received local oscillation optical signals according to the ratio of 10:90 and respectively transmitting the split signals to the receiving end second beam splitter and the receiving end phase modulator; the receiving end second beam splitter is used for averagely splitting 90% of received local oscillation optical signals into two beams and respectively sending the two beams to a receiving end optical rate meter and a receiving end synchronous clock; the receiving end phase modulator is used for carrying out phase modulation on a beam of 10% received local oscillation optical signals and then sending the modulated local oscillation optical signals to the receiving end homodyne detector; the receiving end amplitude modulator is used for carrying out real-time shot noise estimation after the received signal light is subjected to maximum attenuation with the probability being a set value, and sending the result to the receiving end homodyne detector; the receiving end light rate meter is used for detecting light power according to the received signal and uploading the result to the receiving end controller; the receiving end synchronous clock is used for generating a clock signal according to the received signal and uploading the result to the receiving end controller; the receiving end homodyne detector is used for carrying out homodyne detection on the received local oscillator light and the received signal light and uploading the result to the receiving end controller; and the receiving end controller is used for sampling and attack detection according to the received signals.

3. The hybrid attack detection method for a continuous variable quantum key distribution system according to claim 1 or 2, wherein the step S2 is configured to obtain continuous variable quantum key communication data in different attack modes, specifically, obtain continuous variable quantum key communication data in normal communication and obtain continuous variable quantum key communication data in attack; the attacks include single-mode attacks or combined-mode attacks among calibration attacks, low-strength attacks, intercept-retransmit attacks, and saturation attacks.

Images